It is a cheap b.fake pcb.
Hmm... Looking at
the qmk firmware this is a fun one. bfake.h defines how the matrix is laid out:
#define LAYOUT_all( \
K61, K71, K72, K73, K74, K64, K65, K75, K76, K77, K78, K68, K66, K10, K60,\
K11, K01, K02, K03, K04, K14, K15, K05, K06, K07, K08, K18, K16, K20, \
K12, K21, K22, K23, K24, K34, K35, K25, K26, K27, K28, K38, K40, \
K19, K13, K41, K42, K43, K44, K54, K55, K45, K46, K47, K58, K49, K50,\
K09, K00, K39, K30, K59, K69, K57, K29\
And as you can see the second row has Tab, T, Y, { and } as K1x while W, E, R, T and U, I, O, P and ANSI \| are K0x which means they are not all in one row. If this is your board it's very strange that Y, U, I, O, P, { and } all stopped working as that would mean at least two breaks...
There won't actually be a connection from these keys directly to each other but you should be able to get continuity on all the switch to diode connections (like the red lines above) and across all the diodes (like the orange line) If a switch isn't connected to it's diode you can solder a wire between them to fix it but hopefully you won't need to.
To confirm what we're dealing with please can you check the diode connections (orange line) on a row which is working - either all the diodes will be connected or you'll have the groups of 4/2/4 keys in the letter section as suggested above. Either is fine, just have to be sure what we're dealing with